IN-PERSON Barcelona Livable Cities. Section B

Abstract

Future livable cities rely on sustainable forms of human mobility to mitigate the adverse effects of climate change. This is particularly true regarding human mobility, as individuals seek eco-friendly travel options to offset their carbon footprints. Yet, the absence of convenient and accurate tools for measuring individual carbon emissions by mode and purpose of transport remains a significant barrier. In this paper, we propose a new graph-neural network model, SpeedGNN, that leverages dense GPS trajectories collected through individuals’ smartphones to predict an individual’s travel purposes. In the benchmark experiment, SpeedGNN outperforms traditional deep learning models, such as the long-short-term memory (LSTM) model. Moreover, our model also demonstrates strong flexibility and potential in transfer-learning, achieving high accuracy across geographical regions after fine-tuning with small samples. Lastly, to validate the model’s real-world accuracy, we have developed a smartphone mini-app and, through it, we plan to conduct a multi-day experiment with thousands of participants recruited across China. This mini-app will employ SpeedGNN to infer travel purposes, visualize individual carbon emissions, and encourage behavioral change through a gamified interface. Our contribution? An intelligent individual human mobility analytics and a significant reduction of carbon emissions from eco-friendly travelers.